Tertiary structure assessment at CASP15

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• The Critical Assessment of protein Structure Prediction (CASP) competition evaluates computational methods for predicting protein 3D structures.
• AlphaFold 2 (AF2) demonstrated remarkable performance at CASP14, setting a new benchmark for protein structure prediction accuracy.

Purpose of the Study:

• To report and analyze the results of tertiary structure prediction methods at the CASP15 experiment.
• To assess the performance of various protein structure prediction groups and identify key methodologies.

Main Methods:

• Evaluation of single-chain predictions irrespective of template availability, a novel approach for CASP15.
• Analysis of top-performing groups, focusing on their reliance on AlphaFold 2 (AF2) and other strategies.
• Investigation into the role of deep Multiple Sequence Alignments (MSAs) and variant MSAs in predicting challenging protein targets.

Main Results:

• No single group dominated CASP15; most top performers utilized AF2 in their pipelines.
• Deep MSAs and variant MSAs were crucial for successfully predicting difficult targets, often proteins with few homologs.
• Local prediction inaccuracies correlated with specific structural contexts (interfaces, high B-factors) rather than solely prediction error.
• High-quality predictions were achieved by most groups, proving valuable for experimental structure determination and biological analysis.

Conclusions:

• CASP15 highlighted the widespread adoption and effectiveness of AF2, often in combination with advanced MSA techniques.
• The study provides valuable insights into the strengths and limitations of current protein structure prediction methods.
• High-accuracy protein models generated through these methods significantly aid biological research and discovery.